Sensitivity Integrity for tool monitoring, tuning and matching

In order to maintain the optimal conditions of a tool during production, Fab engineers have to periodically perform offline tool monitors to check its status. Usually, a limited tuning action is required to compensate the time effects that may decay the tool performance. Those tuning actions are basically calculated by a tool sensitivity table which describes the numerical relations between tuning knobs and monitor items. However, the integrity of the sensitivity table may suffer from the differences of tools´/chambers´ actual conditions, such as parts-aging, parts qualities and so forth. The impacts to Cp/CpK can be significant while the SPECs are tightened as technology progresses. A common offline tool tuning technique adopted by Fab engineers is called pre-tune, shown as fig. 1. They use the data from previous monitor to calculate the corresponding values of tool tuning knobs by using sensitivity tables. One or several control wafers are processed with monitor recipes to check the tool status after tuning. The advantage to use this technique is to save the monitoring wafers while the safety of tuning is still guaranteed. The tradeoff is that variation or disturbance happened between monitors will usually be neglected in the calculation process. Moreover, the integrity of sensitivity table used for tuning is concerned due to some hardware related physical phenomena, such as target consumption in PVD causing certain sensitivities to be a function of time, special handles are thus required especially for MIMO (Multi-Inputs-Multi-Outputs) cases in N28 and beyond technologies. We need to design an index to ensure the reliability of using sensitivity tables. The variation or disturbance between monitors has to be evaluated while they have significant impacts to the tuning performances. And also to keep the sensitivity table updated to be correct if it tends to be unfit.